We are regularly asked this question "Do I need Defrag.NSF on all my servers, particularly when they are clustered". The answer is a resounding "YES".
Fragmentation (and "de-fragmentation") does not replicate between databases, but replication nearly always causes fragmentation.
We recently commissioned a new server in our office and joined it to the existing cluster. After creating replicas of the databases on the new server, it is quite amazing to see how much fragmentation the process introduced and the impact it had on performance when reading or backing up those databases (The backup software must read the database first).
The screen-shot below shows the fragmented state of the database replicas on a brand-new Domino server. The other Cluster member-servers hosting the replicas of these databases have had Defrag.NSF maintaining them on an automated schedule, so they all exist on disk in an optimal state as a single fragment.
Utilising the integrated Performance Logging in Defrag.NSF, we can see there was a lot of fragmentation in the databases and there has been a significant improvement in non-cached read performance since Defrag.NSF has run and reduced the fragmentation back down to a single fragment.
Even though the other Cluster members were hosting perfectly defragged replicas of these databases, the fragmented files on this brand-new server would have rapidly deteriorated even further without the required maintenance being performed by Defrag.NSF.
You can find out more about Defrag.NSF at www.preemptive.com.au